[
https://issues.apache.org/jira/browse/HDFS-7317?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14192494#comment-14192494
]
Andrew Wang commented on HDFS-7317:
-----------------------------------
One issue with how StoragePolicies are currently stored is that we reserve some
bits for a policy ID in the INode header, and then that directly maps to a set
of StorageTypes (with fallbacks). When we add the additional dimensions of EC,
compression, and how each of those are defined, it seems hard to map all the
possible combinations to a single flat id space. It might be better to use a
more complex datastructure instead.
> Rename StoragePolicy
> --------------------
>
> Key: HDFS-7317
> URL: https://issues.apache.org/jira/browse/HDFS-7317
> Project: Hadoop HDFS
> Issue Type: Improvement
> Affects Versions: 2.6.0
> Reporter: Andrew Wang
>
> As discussed on HDFS-7285, StoragePolicy might not be the best name for what
> StoragePolicy currently is, which is a hardcoded mapping to a StorageType.
> Ideally the "policy" is what determines the data temperature in the first
> place, with the temperature then mapping to the actual StorageTypes to use.
> There were a number of suggestions presented, e.g. StorageTag,
> StoragePolicyTag. Let's figure this out here.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
